🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
Desktop manager for AI agents and configs
BootAgent scans a machine for supported agents, providers, skills, and MCP servers, then lets you configure and launch them from one local app. It keeps profiles, imports, exports, and backups in one place so you do not have to edit several config files by hand.
Builders who want to keep Claude Code, Codex, and other agent setups in one local place.
You can set up and move agent workflows with less manual config editing.
What it does
Detect and manage agents
Finds supported CLI and desktop agents already installed, then lets you install, update, launch, or uninstall them.
Connect providers and profiles
Supports built-in providers and compatible endpoints, with reusable profiles and protocol checks before writing config.
Manage skills and MCP servers
Keeps a local skill library, discovers MCP servers from initialized installs, and lets you choose sync targets before applying changes.
Import and export setups
Moves providers, profiles, selected MCP servers, and skills between machines with JSON and bundle formats, plus preview and rollback.
Marketplace browsing
Lets you search skills, MCP servers, plugins, prompt collections, and workflow templates from bundled and online sources.
Local recovery and backups
Stores local backups and history versions for configs and can restore or roll back changes after imports.
How to get it
- 1The shortest first-run path is
Launch BootAgent → Choose or add a Provider → Create an API Key in that Provider's official account page → Choose an Agent and a Profile → Run setup → Send a first request
- 2For users, the release packages above are the supported installation path. For…
git clone https://github.com/MaimoryLab/BootAgent.git cd BootAgent cd frontend pnpm install --frozen-lockfile pnpm run build cd .. go run -tags wails ./cmd/bootagent-desktop
- 3For a production binary on the current host, use task build:desktop. Useful checks are
go test ./... go test -race ./... go vet ./... cd frontend && pnpm run test && pnpm run build
README
BootAgent
Many Agents. One place to manage them.
BootAgent is a local desktop workspace for AI coding Agents. It helps you detect what is already on your machine, connect a Provider, configure an Agent, and get to a first successful request without editing several tool-specific files by hand.
Start here
| You want to... | Start with... |
|---|---|
| Try an Agent for the first time | Download a release, then follow the AI Agent Kit |
| Keep existing Agents in one place | Open Environment overview and let BootAgent scan the machine |
| Find a Skill, MCP server, plugin, or AI product | Open Marketplace, search or filter, then open an item for its details and upstream links |
| Move a setup to another machine | Open Settings > Import and export and choose a v1 JSON file or a v2 bundle |
| Fix an install or update | Open Task Center for the source, progress, and diagnostic output |
The shortest first-run path is:
Launch BootAgent
→ Choose or add a Provider
→ Create an API Key in that Provider's official account page
→ Choose an Agent and a Profile
→ Run setup
→ Send a first request
What you can do
Manage Agents
- Detect, install, update, launch, and uninstall supported CLI and desktop Agents.
- Detect Agents installed outside BootAgent when their command or known installation path is available.
- See long-running install and update tasks in Task Center, including progress, source, and cancellable steps.
- Uninstall selected installation instances while keeping Profiles, Providers, configuration files, and conversations.
Connect Providers and Profiles
- Use built-in Providers or add an OpenAI-compatible or Anthropic-compatible endpoint.
- Check an endpoint using the protocol that the selected Agent actually uses.
- Save reusable Profiles and select the Profile and model from an Agent's configuration screen.
- Keep API Keys in private local storage; they are not placed in ordinary summaries or recommendation prompts.
Manage Skills and MCP servers
- Keep a local Skill library, scan Skills already present on supported Agents, and choose which Agents receive each Skill.
- Discover MCP servers from initialized Claude Code, Codex, OpenCode, Kilo CLI, and Hermes installations.
- Store MCP servers in a local user-level Registry, select sync targets explicitly, and apply changes only when you are ready.
Local utilities
- Bootstrap Node.js, uv, and Aider's managed Python runtime when an Agent needs them.
- Optionally run local API format conversion and launch BootAgent at login; both features are off by default.
- Migrate existing Codex and ChatGPT Desktop conversations into BootAgent's local provider bucket. This migration intentionally creates no history backup.
Discover tools in Marketplace
- Browse Skills, MCP servers, plugins, standalone AI products, prompt collections, and workflow templates.
- Combine category, source, use-case, API-key, and multi-type filters. An item can belong to more than one type.
- Refresh supported public sources independently. A source failure keeps the bundled catalog available instead of blanking the market.
- Open a detail page for the description, install guidance, README when the source provides one, and upstream links.
- Ask an installed Codex or Claude Code CLI to shortlist tools for a stated need. Returned IDs are checked against the current catalog before display.
- Save recommendation requests and result snapshots locally so a previous search can be reopened later. Recommendation history is never uploaded as telemetry.
Keep setups portable
- Import and export Providers, Profiles, selected MCP servers, and Skills.
- Legacy v1 JSON files remain importable. The v2 portable bundle separates configuration JSON from a Skills ZIP.
- Skills first enter BootAgent's managed library; you choose target Agents from the Skills page afterward.
- Preview new, overwritten, and conflicting resources before applying an import. Writes use snapshots and rollback on failure.
- API Keys and MCP secrets are omitted by default. Password-encrypted export or explicitly confirmed plaintext export is available.
Supported Agents
BootAgent supports detection, configuration, launch, or installation guidance according to each Agent's upstream contract. An entry below does not mean that BootAgent redistributes the Agent package; the detail page shows the available install source.
| CLI and local Agents | Desktop Agents |
|---|---|
| Codex · Claude Code · OpenCode | DSH Desktop · Claude Desktop |
| Kilo CLI · Aider · OpenClaw | ChatGPT Desktop · WorkBuddy |
| Hermes Agent · Kimi Code · Pi | WorkBuddy AI · ZCode |
| DeepSeek Harness (local web app) |
Claude Desktop can be detected and launched on macOS and Windows, but it must be installed by the user from the official download page. BootAgent writes compatible configuration and does not download or proxy Claude Desktop.
DeepSeek Harness can use its shipped DeepSeek route or a configured compatible Provider. It opens a local web app after installation; its own onboarding commands remain the source of truth.
JieKou.AI, PPIO, Novita, DeepSeek, and Moonshot are built-in Providers. Custom OpenAI-compatible and Anthropic-compatible Providers can be added from the Provider page.
Download
Download the latest package from GitHub Releases.
| Platform | Recommended package | Architectures |
|---|---|---|
| macOS | DMG | Intel and Apple Silicon |
| Windows | NSIS installer | x64 and ARM64 |
| Linux | AppImage, deb, or rpm | amd64 and arm64 |
Linux also provides an OTA ZIP archive. Release artifacts include SHA256SUMS. macOS packages are Developer ID signed and notarized. The release channel is still a Wails Alpha technical preview: Windows and Linux packages are currently unsigned and may require an operating-system approval on first launch.
BootAgent does not redistribute Agent packages and does not bundle Node.js, Git, WebView, or API Keys. Missing prerequisites are reported with a link to the relevant official installation instructions.
Data, privacy, and recovery
- BootAgent is local-first. Provider credentials, configuration, recommendation history, and backups stay on the machine by default.
- Recommendation prompts contain the stated need and public catalog metadata only; the recommendation process receives no install or file-write tools.
- The latest three historical versions are kept for each Profile, Provider, MCP, Agent configuration target, and Skill by default. Backups live under
~/.bootagent/backup; retention can be changed in Settings. - Uninstall removes the selected program instance, not the user's Profiles, Providers, configuration files, or conversations.
- BootAgent is not a VPN, proxy, shared-key service, or Agent package distributor. Downloads use official sources, authorized mirrors, or documented manual installation paths.
Troubleshooting
| Symptom | First check |
|---|---|
| Agent is not listed | Refresh Environment overview, check that the command is on PATH, or use the Agent's documented manual install path. |
| Provider connection check fails | Confirm the endpoint and model match the Agent's protocol; an endpoint that exposes only another API is rejected before writing configuration. |
| Marketplace looks incomplete | Check the source status and refresh. A bundled snapshot remains available when an online source is unavailable. |
| Import reports a conflict | Review the preview, choose skip or overwrite, and apply only after checking the affected resource names. |
| BootAgent cannot check for an update | In Settings, switch between GitHub and the domestic Gitee mirror, then retry. Existing installation files are not replaced by a failed check. |
| The Agent will not download | Follow the official install page or install manually, then return to BootAgent to detect the local installation. BootAgent does not bypass network restrictions. |
Build from source
For users, the release packages above are the supported installation path. For development, you need Go (the version declared in go.mod), Node.js, pnpm 11.21.0, and the target platform's WebView dependencies. Linux builds use GTK4 and WebKitGTK 6.0.
git clone https://github.com/MaimoryLab/BootAgent.git
cd BootAgent
cd frontend
pnpm install --frozen-lockfile
pnpm run build
cd ..
go run -tags wails ./cmd/bootagent-desktop
For a production binary on the current host, use task build:desktop. Useful checks are:
go test ./...
go test -race ./...
go vet ./...
cd frontend && pnpm run test && pnpm run build
Project links
- AI Agent Kit: guided setup for a first request
- Documentation: specifications and architecture decisions
- Security policy
- Contributing
- Public site repository
- Issues and feature requests
Star History
Sponsors
BootAgent is released under the Apache License 2.0. Third-party notices are listed in NOTICE.
Files in the repo
- .claude
- .github
- brand
- build
- cmd
- docs
- frontend
- internal
- manifests
- scripts
- third_party
- .gitignore
- .markdownlint.yml
- AGENTS.md
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- go.mod
- go.sum
- LICENSE
- manifest_embed.go
- NOTICE
- README_ZH.md
- README.md
- SECURITY.md
- staticcheck.conf
- Taskfile.yml
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More tools
The best-benchmarked open-source AI memory system. And it's free.
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.